{ "model_type": "mistral", "architectures": ["MistralForCausalLM"], "vocab_size": 32000, "hidden_size": 4096, "num_hidden_layers": 32, "num_attention_heads": 32, "num_key_value_heads": 8, "intermediate_size": 14336, "max_position_embeddings": 32768, "rms_norm_eps": 1e-5, "rope_theta": 1000000.0, "torch_dtype": "bfloat16", "attention_dropout": 0.0, "hidden_act": "silu", "sliding_window": 4096, "bos_token_id": 1, "eos_token_id": 2, "pad_token_id": 32000 }